Professional Documents
Culture Documents
Radoslav M. Dimitrić
arXiv:1305.5173v1 [math.HO] 21 May 2013
dimitricr@member.ams.org
1 The first version of this paper was presented at the joint American Mathematical Soci-
ety/Mathematical Association of America Meeting, Boston, January 4, 2012. The paper was
presented at the 12th International Congress on Mathematics Education (ICME12), Seoul, South
Korea, July 8-15, 2012.
1
2
concepts, that in turn, depend on precise, formal definitions of notions of two and
higher dimensions (and representations in coordinate systems). Perhaps a notion
close to this aspect is “concept definition” as discussed in Tall & Vinner (1981), Vin-
ner (1991), Vinner & Dreyfus (1989). However these works concentrate primarily
on “concept image” which is undoubtedly important, however the metaphors (such
as in “concept images”) cannot be made operational and stable, unless there is
definite command of relevant and definite definitions these metaphors should map.
By metaphor, I understand the classical (and somewhat modernized) definition of
metaphor originating with Aristotle in his On the Art of Poetry and Rhetoric “car-
rying of expression from one object onto the other...” In short, metaphor consists
of three parts: the first object, the “carrying” i.e the process of connecting the
two objects, and the second object.” Metaphors are different if at least one of their
three parts differs from the corresponding part of another metaphor.
The present paper gives an example of analysis of students tests in a default
classroom setting. Among other things, I make suggestions as a practical guide in
teaching the subjects involved, but I also point out areas worthy of further education
research. Notable studies exist with related themes and I mention here Selden et
al (1989, 1994, 2000) where the authors test (outside of the classroom setting) self-
selected sample of students (after they complete the usual calculus sequence) on
routine and non-routine problems.
Seven concrete test problems are presented here (to be completed in 100 minutes
– the time that is perhaps tight to complete all the problems). By departmental
decision, graphing calculators are allowed, starting from the most elementary re-
medial courses to post calculus courses. My comments accompany each individual
problem with final conclusions at the end of the paper. Students’ solutions are
quoted verbatim, without grammatical corrections.
When making the exams, I try to adhere to some extent (and to the extent
possible in classes at this level) to Bloom’s taxonomy (Bloom (1956), and its revi-
sion, Anderson et al (2001)) and devise exam questions of different difficulty to test
various aspects of students’ knowledge. The issue of the difficulty level and all the
ramifications is a complex issue. I expound on this topic in more detail in Dimitric
2012b, while adopting here what is possibly the crudest (but still useful) graduation
in Bloom’s two-dimensional taxonomy array. The first two problems were meant
to be of the simplest kind and students’ scores on the problems show that my gage
of relative difficulty was correct. This however does not mean that students’ per-
formance was appropriately good, as much as I expected or hoped for – the mean
scores on the individual problems, out of maximum 15, were as follows (Problems 1
to 7, in order): 8.04, 6.39, 2.96, 3.52, 0.96, 2.22, 3.39 while the respective medians
were: 8, 7, 1, 3, 0, 1, 3. My rating of problem 5 was not the highest difficulty,
4
Problem 1. Given three points in 3D space: A(2, 0, a), B(0, 1, a), C(−1 − 1, 0),
where a is an arbitrary given constant.
a) Calculate the vector/cross product AB~ × AC.
~
b) Calculate the scalar/dot product AB ~ · AC
~
c) For what values of a are vectors AB~ and AC~ (i) perpendicular, (ii) parallel?
d) Write an equation of the plane that contains points A,B,C.
Normally, analytic geometry is not a part of the third calculus; rather, it is often
“reviewed” in previous calculus courses, and at any rate is included in most “coffee
table” calculus textbooks. My department decided to include it in calculus III as
a “review” and rightly so, given the level of mathematical skill of students taking
such courses. There is however a drawback in this kind of remediation since the
topics that are left for the end of the course are sometimes, if not half of the time
(depending on the class), left out for sheer lack of time. But those topics are (or
should be) the core of calculus III – Green’s, Stokes’ and divergence theorems. This
course is fairly standard in the US – it includes topic on vector-valued functions
and surfaces, partial derivatives, optimization and Lagrange multipliers, double and
triple integrals and their properties and techniques of evaluation, vector fields, line
integrals, surface integrals that all culminate in Green’s, Stokes’ and Divergence
theorems.
2 The issue of use of more pre-calculus in calculus was addressed in Dimitric (2001); that
approach however requires a greater sophistication than is exhibited from a typical calculus student
3 Selden et al report that their routine stage is well fulfilled by the students they test, however
their testing was on the material several courses earlier than the course levels the tested students
were at, at the time of testing
5
This problem was meant to test the elementary 3D concepts at the level of def-
initions and simple steps. I have used a constant a in order to prevent “button
pushing” on programmable calculators. This causes some problems to some stu-
dents who still think that a constant must be given as a concrete decimal number
(most often an integer), not a letter, which (according to many students) is an
unknown variable, by default. This kind of problem could be eliminated if there
were beginning courses that would teach basic mathematical culture, including the
concepts of a constant and variable; in fact such courses should be made mandatory,
at least for mathematics majors.
A number of students stumble upon an objective problem, namely the fact that
we represent vectors both by their end-points (that are each ordered triples) as
well as by a single triple of vector’s coordinates (that are the difference of those
end-points). This is an objective confusion since localization and affinity of 3D
space make this “gluing” possible. The instructor should thus spend some time
in elaborating the concept of vector translation which produces the same vector
(strictly an equivalent vector) that begins at the origin. And judging by further
errors made, the instructor should also make sure that students remember that you
need to subtract the initial point from the end point of the vector to get the vector’s
3D components: AB ~ = (−2, 1, 0) and AC ~ = (−3, −1, −a).
One idea to avoid this confusion is perhaps to use round brackets for ordered
triples representing points and to use angle brackets to represent vectors, however
I did not find that this alone can remedy the problem.
The errors even at this first step are in simple arithmetic (subtracting greater
number from smaller), mishandling of minus signs and parentheses, multiplications
by 0 treated as multiplications by 1, and getting vectors to be the opposite of
what they should be. Arithmetic errors come somewhat from the exam pressure
and to a greater extent because of addiction to calculators that result in students
being unable to do even the simplest arithmetic operations in their heads. I have
mentioned elsewhere [Dimitric 2003, 2009] that students did a bit better when they
were not allowed calculators into the exams.
One student “finds” the vectors by dot multiplying their endpoints, which is in
the domain of confounding (with the dot product). I wanted to check that they
know definitions of the cross (vector) and the dot (scalar) products and how to
compute them if vectors are given in the component form. In the scalar product
all that is required is component-wise multiplication and addition, but the vector
product involves a bit of a twist in that students have to either use the appropriate
determinant or express the vectors through the basis vectors i, j, k and use proper-
ties of cross multiplication. A majority of the students used the determinant and
perhaps it may be due to the fact that I used the determinant more often than other
methods in computing the product when working out exercises at the blackboard.
That should then imply that including more computations with vectors spanned
by the basis may be helpful.
In calculating the cross product determinant (which computes to be −ai − 2aj +
6
5k), the usual arithmetic errors occur, now including calculation of the three by
three determinant. Not all students remember the determinant definition of the
cross product. The student that “found” vectors by multiplying their endpoints
also formed a 2 by 3 “determinant” and did some inventive multiplications inside
to get his incorrect result.
Not all students remember the coordinate definition (computation) of the scalar
product (equal to 5 in this case). Here is how one student does it: (−2, 1, 0) ·
(−3, −1, −a) = (6, −1, 0).
Part c) was meant to check whether students learned the simple connection be-
tween the two products with vectors and orthogonality and parallelism. Some con-
fuse the two and relate parallel to AB~ · AC
~ = 0 and perpendicular to AB ~ × AC~ =0
(many may miss the fact that the latter 0 is a vector). Here the instructor has more
work to do in devoting more time in explaining the distinctions between the two
products, such as that one is a scalar and the other is a vector and spend more time
on the geometric definitions of the two (in addition to their consequent computa-
tional definitions). One may add mnemonic rules to aid the students to remember
which is which and how they are related to these important geometric conditions.
Another student confuses perpendicularity with the condition for perpendicularity
of two lines; he states: “parallel when a = 0. Can’t be perpendicular because a cant
[sic] become a negative reciprocal...” That a should be 0 for parallel or perpendicu-
lar is frequent. One student computes the dot product correctly to be 5 and states
correctly the condition of perpendicularity, but claims that the two vectors are al-
ways perpendicular; this could be simply a lack of concentration on the student’s
part? Another student “solves” this as follows: “AB ~ and AC~ parallel if they point
in same or opposite direction, since AB ~ = h−2, −1, 0i and AC ~ = h−3, −1, −ai, AB ~
~
and AC are not parallel because first component is unequal.” Perhaps one needs to
spend more time on the concept of parallelism by way of one vector being a scalar
multiple of the other, resulting in all corresponding coordinates being proportional.
There were also students who overcomplicated, at least with some of their solu-
tions. Calculus II or III students often take linear algebra concurrently and are only
too eager to demonstrate what they are exposed to in that class, Gauss elimination
being one of the favorites, since considerable attention is paid to it in linear algebra
courses and it is a fairly algorithmic procedure that students always favor. Here is
how one student approaches questions in part c) (decomposition into perpendicular
and parallel components of a vector is also in his mind, for it may have been a part
of a review session before the exam): i) v⊥ + v|| = ||v||, a = −5, ii) a = 2.25. To get
this a, he does Gauss elimination in the matrix made up of rows AB ~ and AC ~ and
gets matrix with rows (1, −1/2, 0) (incorrect) and (0, −5/2, −a) and then matrix
with rows (1, −1/2, 0) and (0, 1, 2a/5) and then rows (1, 0, 9a/10) and (0, 1, 2a/5),
then equates 9a/10 = 2a/5 and gets 45a = 20a from which he derives a = 2.25.
Ending up with an incorrect answer may be prevented, with some chance of success,
by the instructor telling students to check their answers by plugging in the values
they get.
7
This exercise checks basic properties related to partial derivatives and also whether
students know the conditions, namely that ∂F1 /∂y = ∂F2 /∂x and that the compo-
nents should have continuous partial derivatives. If these conditions are satisfied,
then they should also demonstrate whether they can find such a potential func-
tion f . To simplify considerations on smoothness of partial derivatives, I made the
assumption on the existence of all the partial derivatives which then imply their
continuity. If such an f exists, then, we have to have
∇f = (∂f /∂x, ∂f /∂y) = (y 2 , x), i.e. ∂f /∂x = y 2 and ∂f /∂y = x,
consequently ∂ 2 f /∂y∂x = 2y, ∂ 2 f /∂x∂y = 1. Since we are assured that all partial
derivatives exist and are therefore continuous, a Clairaut’s theorem asserts that
these two mixed second order partials must be equal (which does not happen here,
apart from the points on a single horizontal line 2y = 1).
Students also had a ready-made result that was used in a few exercises in class,
namely that if (F1 , F2 ) = ∇f is a gradient vector field (and f is a potential func-
tion) then one must have ∂F1 /∂y = ∂F2 /∂x, which would again lead to the same
conclusion of no solution.
However, the mathematical results on conditions and existence are not always
welcome or easily adopted by science students; they prefer to “do something” – com-
pute a number, a value of a function or such (which may be classified as an “action
conception of function”, Dubinsky & Harel (1992)). Thus, some students can ignore
conditions and go straight to solving the system: (1) ∂f /∂x = y 2 , ∂f /∂y = x (2).
After integrating the first equation with respect to x one gets f (x, y) = y 2 x + φ(y),
hence ∂f /∂y = 2yx + φ′ (y), which again leads to non-existence of the solution
(2y = 1).
Indeed students’ solutions (or incorrect solutions) show that they mostly at-
tempted to find the solution by solving the system of the two pde, by finding
8
one student makes (after finding the anti-derivatives) is that he writes f (x, y) =
(xy 2 , xy) thus confusing the roles of the players in the question. Knowing that the
codomain of f is one-dimensional would have prevented that error and that, yet
again would have been clear if the student wrote all three parts of the function:
f : R2 −→ R.
The purpose of this exercise is to test students on the routine of the chain rule
in several variables (three), each of which depends on three variables. Instruction
was geared towards showing that multi-variable chain rule is actually a natural
extension of the one variable case and that, for that matter, the chain rule formulas
∂f ∂f ∂x ∂f ∂y ∂f ∂z
= + + (1)
∂r ∂x ∂r ∂y ∂r ∂z ∂r
∂f ∂f ∂x ∂f ∂y ∂f ∂z
= + + (2)
∂θ ∂x ∂θ ∂y ∂θ ∂z ∂θ
are understandable in view of the one variable chain rule. Just writing down these
formulas earned them 1/3 of the points for this problem. They also had to know
the cylindrical coordinate formulas, from which then they had to find the needed
partial derivatives
∂x ∂y ∂z ∂x ∂y ∂z
= cos θ, = sin θ, = 0, = −r sin θ, = r cos θ, =0
∂r ∂r ∂r ∂θ ∂θ ∂θ
(this would earn the student almost another third of the points). They would be
ready to substitute these partials into (1) and (2) to get the final answer:
∂f ∂f ∂f ∂f ∂f ∂f
= cos θ + sin θ; = (−r sin θ) + r cos θ.
∂r ∂x ∂y ∂θ ∂x ∂y
A number of errors appear here. A few students in fact do not quite know the
cylindrical coordinate formulas yet (Bloom’s step one). Confusions are with the
polar coordinates in two dimensions x = r cos θ, y = r sin θ (omitting the third
equation z = z, or writing z = r); this is also confused with the equation of the
right circular cylinder x2 +y 2 = r 2 or even = z 2 , since the thinking is that z = r. At
least one student confused cylindrical coordinates with spherical ones, undoubtedly
attributed to the fact that, in a relatively short time, both of these coordinate sets
were used in various problems and topics covered in class.
10
Problem 4 Given points A(0,1), B(1,2) and C(3,0) (in 2D) and the curve γ that
consists of the portion of the parabola y = 1 + x2 between points A and B and the
straight line segment connecting B and C, find the line integral
Z
~ · d~s
F
γ
where F~ (x, y) = (x2 − y, y 2 + x) and the curve is traversed in the direction from A
to B to C. What would be the value of the integral if the orientation of the curve
γ were from C to B to A?
This one is meant to test whether students have gotten basics of the line integral
of a vector field. Parts of the solution(s) that were assigned partial credit were:
Drawing the graph of the path and recognizing that the integral overR γ is Rthe sum R
of the two pieces – over the parabolic path and over the line path: γ = γ1 + γ2
(2 points). Then the students should know how to parametrize the two paths,
for instance: γ1 : x(t) = t, y(t) = 1 + t2 , t ∈ [0, 1] and γ2 : x(t) = t, y(t) =
−t+3, t ∈ [1, 3]. One of course needs to know the basic technical integration formula
Rb
~ ~ y(t)) · (x′ (t), y ′ (t))dt and to find correct derivatives
R
γ
F(x, y) · d~
s = a
(F(x(t),
x′ = 1, y ′ = 2t, for γ1 and x′ = 1, y ′ = −1 for γ2 (3 points). Routine polynomial
~ s = 2 and ~ s = 0 (6 points), so
R R
definite integration will then produce γ1 Fd~ γ2
Fd~
R
that γ = 2 + 0 = 2 (2 points). The orientation question tests whether a student
12
knows that RreversingR the path orientation changes the sign of the value of the
integral i.e. γ − = − γ + (2 points).
Apart from the usual arithmetic and algebraic errors, some students are unable
to sketch the (entire) integration path (one student attempts to draw it in 3D,
undoubtedly influenced by a good number of 3D images of surfaces and curves
done on other topics). Ability to draw these simple curves should be a prerequisite
for taking any of the calculus courses (see here Dimitric (2012a)). Again, some
additional time should be spent on functions defining curves and surfaces in different
dimensions; it is highly recommended that functions are presented not only as
formulas, but as ordered triples consisting of domain, codomain, and the formula(a
rule of correspondence). Analysis of the dimensions of the domain and codomain
would then aid the student in recognizing the context as to whether he is dealing
with a surface or a curve and whether it is two, three, or more dimensions. Striving
to make students switch from action to process conception of function would not
help, unless students actually were well familiar with the complete functions, namely
domain, assignment rules and codomain, or luckily the range.
There were a couple of students who only drew the graph of the path and did
nothing else. When it comes to parametrizations the mistakes are in not parametriz-
ing the line or the parabola piece, or not determining correct limits of the parameter
in either of the two pieces. Some students however see no need for parametriza-
tions and use the Cartesian coordinates of the two paths (and some may have the
line segment Cartesian equation incorrectly). This is because those students likely
confound this with the double integral exercises they have been exposed to. Prob-
lems with parametrizations should be diminished by the instructor’s explanation
why parametrizations are needed, as well as working out some simple but also
some intricate parametrizations, while elaborating on the conditions, advantages
or disadvantages of one parametrization over the other.
A number of students then use an incorrect formula to integrate this vector field,
~
namely γ F · d~s = F (c(t)) · ||c′ (t)||dt, which is a formula for the line integral of
R R
a scalar function F . Yet again, dimensional analysis here is of great help to decide
the correct context for solving the problem. A list of “invented” formulas for the
integral is as follows:
R1 2 R1 R1
0
(x − y, y 2 + x)ds = ( 0 x2 − yds, 0 y 2 + xds);
R1
~ ds + 3 F~ ds;
R
F
R02 R 2 1
2
R3R2
0 1
(1 + x − y)dydx + 2 0 (−x + 3)dydx = . . . = 7 + 17 = 24;
R0
1+x2
(x2 − y, y 2 + x), f~(x, y) = (x3 /3 − y 2 /2, y 3 /3 + x2 /2) from C to B to A.
R 1+x2 2
−x+3
(x − y, y 2 + x) and stops here.
R3 2
(t −(1+t2 ), (1+t2 )3 +t)· ∇F ~ (x, y) = (this comes after a correct parametriza-
0
tion of the parabola path, but no parametrization of the line path, and correct
determination of ∇F (x, y)). The incorrect answer obtained was 71187/70
~ ds = b (c(t)2 −c(t), c(t)2 +c(t))·(c′ (t))dt These answers are a good indication
R R
c
F a
13
that more work needs to be done in the area of vector field integral, certainly more
examples to illustrate differences in integration of a vector field vs. scalar function.
One student has the following incorrect solution: He draws the graph fairly cor-
rectly, and seems to parametrize the portions accordingly, but instead of evaluating
the integral in the two portions, he “adds” the portions of the path as follows:
C1 = y = 1 + t2 + C2 = y = −(t − 3), c(t) = C1 + C2 = 2y = t2 − t + 4, writes
c(t) = (t2 −t+4)/2 and c′ (t) = (2t−1)/2 and then F (c) = (t2 −(t2 −t+4)/2, (−(t2 −
t + 4)/2)2 + t) and F (c) · c′ = (t2 − (t2 − t + 4)/2, (−(t2 − t + 4)/2)2 + t) · (2t − 1)/2
R1
and then (2t − 1)/2 · −1 (t2 − (t2 − t + 4)/2, ((t2 − t + −4)/2)2 + t).
Most students answered correctly the orientation question, but there were a
couple of incorrect answers as follows:
“The value wouldn’t change going the other direction since the line didn’t change
shape or orientation.”
”The value of the integral of the orientation of curve γ went from C to B to A
does not change as if it went from A to B to C.”
Here it would be worthwhile to spend extra time time with the one-dimensional
case and why the sign of the integral changes when the orientation of the interval
of integration is reversed.
There are also some semi-correct answers, as follows:
“If the orientation was from C to B to A the value of the integration would be
negative” (I corrected this to “opposite”).
This comes from the universal error (colloquialism) that students make, namely
identification of “opposite” with “negative” which is not an innocent error, but is
rather a linguistic sloppiness impeding students’ understanding of concepts, and an
effort should be made to correct it.4
Problem 5 The function f (x, y) is defined as follows on the domain [0, 1]×[0, 1]:
x
, if y 6= 0
arctan(1/y 2 )
f (x, y) =
2xπ, if y = 0
Find all the points where the function is continuous (or, complementarily, where
it is not continuous) in its domain.
the paper), is not very useful in functions of several variables. I have emphasized
on numerous occasions that the definition of continuity of f at a point P is that
limx→P f (x) = f (P ). Although this is not such an overly complicated definition
students do not readily accept it, for the simple reason that it is not the instant
product they like – finding limits may be an involved process and may look as a
roundabout way to establishing what they like to think of as the intuitive concept
of continuity.
Certainly it would be nice to check that the function is defined in the proclaimed
domain, by noting that arctan(1/y 2 ) never turns into 0. The first stage of the proof
would be to note that for points that have non-zero y-coordinate, the first part of the
definition of f is the quotient of two continuous functions. Students are expected to
say that x is continuous as a polynomial and that a rational function with non-zero
denominator is continuous as a composition of two continuous functions arctan and
the rational function 1/y 2 . This argument so far would have earned them a good
partial credit. Then they need to check continuity at points of the form (a, 0) i.e.
they need to check that lim(x,y)→(a,0) f (x, y) = f (a, 0) = 2aπ. Students would have
to find the lim(x,y)→(a,0) x/ arctan(1/y 2 ) = a/(π/2) = 2a/π. This means, that in
order to have continuity at such points, the equality 2aπ = 2a/π must be satisfied
and that happens only for a = 0. Hence the conclusion is that f is continuous at
all points (x, y) for which y 6= 0, also at (0, 0) and discontinuous at all other points,
namely all (a, 0), where a 6= 0.
Here is how one student (the weaker of the two high school seniors who took the
class) attempts to solve the problem: She first rewrites the function in “her own”
way: tan−1 (1/y 2 ) and then further “simplifies” the function into x = tan(1/y 2 ) · y,
where the last y stands for the function, namely f (x, y). My experience helps me
to foresee this kind of error, so I warn the students in classes that it is safer to
use arctan, rather than tan−1 lest they mistake it for the reciprocal, rather than
the inverse function. However one has to spend additional time in clearing up this
notational confusion. Almost all widely used calculus textbooks (see e.g. Rogawski
(2008) or Stewart (2012) unwittingly or purposefully use the inverse function nota-
tion, lest they appear not to be sufficiently “modern.” It is as if we should be writing
square−1 to denote the square root function, just so we can show it is inverse of
the basic quadratic function. The same bad habit is seen with calculator keys that
almost exclusively use the inverse function notation. This notation is certainly
useful in general discussions about inverse functions, but is a bad notation when
names for inverse functions exist (for instance the arc functions such as arctan),
which in their names have interpretation, if not the meaning of a particular inverse
function. It has been noted many times before that good notation is golden and in
this case it would result at least in students not confusing the inverse function with
the reciprocal of the function.5 To follow, the same student further draws that part
5 Most
students are genuinely and thoroughly surprised to learn that tan−1 does not denote
the reciprocal of tan. It is an interesting functional equation to find all invertible functions f
whose inverse equals its reciprocal.
15
of the square root function graph (in the 2D Cartesian coordinate system) that is
in the unit square and claims it is the graph of x/ arctan(1/y 2 ) – all graphs for
her are in two dimensions. Then continues on to say “derivative of ” f ′ (x, y) =
substitution u = tan−1 (1/y 2 )... d2xπ/dx = 2π!! [sic]. Then claims that the func-
tion is not continuous for x = 1 and concludes: “The graph x/ arctan(1/y 2 ) lies
in the positive (x, y) plane and the function is discontinuous at x = 1.” Yet again,
more work must be done simply on functions of more than one variable, in what
their domains and codomains are, as well as their graphs, for, even if students have
better understanding of functions than the action conception, without firm grasp of
domains and codomains or ranges of participating functions would not lead them
to correct solution.
Another student evaluates the function at the four corners of the domain and
concludes that “the function F (x, y) is continuous throughout its domain.” This
comes from students’ confusion with procedures that were used in class when finding
absolute extrema where points on the boundary of the region (or at the corners of
polygonal domains) were crucial. This is confirmed by another student finding the
partials” fx , fy (that seem to be incorrectly done “partial” anti-derivatives), which
he then sets equal to zero and solves the resulting equations. Unquestionably, these
confusions point to insufficient time that students had in digesting the material,
be it objectively (one semester for all the advanced material) and/or subjectively
– insufficient work on the students’ part).
One student correctly computes the function values at the corners, then writes
−x2
(
2 tan−1 (y 2 )−π ,
fx =
2
x π
−2 2 tan−11(y2 )−π ,
fy =
2xπy
He then solves when the parts of these “partials” are equal to 0, conveniently
getting x = 0 as a solution to the first system and not getting anything (“false”)
for the solution of the first part of the second expression; however, from the second
part of the second system he gets xπy = x, y = 0 and says “these points are on
the boundaries.” This is a confusion with the problem of finding the extrema of
functions; the conclusion is that “the function is continuous on the points (1,1) and
(1,0) and not continuous at (0,0) and (0,1).”
One student attempts to find the double integral, over the square, of the first
defining part of the function and, after some incorrect substitutions, etc., arrives at
an incorrect value of the integral. Then he states, as a matter of fact that: “It is
continuous because it is defined everywhere and limits exist.” Double integrals are
frequent among the incorrect solutions here.
Some students know the correct definition of continuity, but cannot find the limit
of the function at a general point, rather they try to find it in a corner point or
16
two, or invent their own limit rules. Thus one states: “2xπ is not dependent on y
and is a continuous function that will be well-defined within the domain” and then
x 1
“ lim 2
= lim f (x, y)′ = .”
y→0 arctan(1/y ) y→0 limy→0 arcsec(1/y 2 )
One student discusses continuity of the parts separately: “For f (x, y) = 2xπ
this is continuous for all points work from the domain x= all real numbers. For
x/ arctan(1/y 2 ) all points work except for 0 from the domain. Because only way
x/ arctan(1/y 2 ) is not continuous is if the denominator is equal to zero. Only value
it will equal zero is if y = 0. So all points on the function f (x, y) is continuous
except when y = 0.” This student did receive partial credit, but here comes again
the problem of functions, in that these piecewise definitions of functions are often
seen as two separate and independent functions, partly because it is more convenient
to do so. Breidenbach, Dubinsky, Havks, Nichols (1992) use “split” functions to test
whether students’ understanding of functions is the process one. This is something
that needs to be addressed in pre-calculus or the beginning algebra courses.
One student states: “... a function is not continuous where it is undefined,” then
goes on to find correctly that the denominator in the fraction part is never 0 and
concludes that there are no points of discontinuity. This statement is an innocent
regurgitation of classroom and textbook folklore where continuity of a function at
a point is still considered even though there is no function (i.e. the function is not
defined there) – the point addressed in Dimitric (2004); see furthermore, Rogawski
(2008) and Stewart (2012) where the instructing examples and the exercises on
“continuity” are almost exclusively such that a student finds zeros of the denomi-
nators of rational functions and pronounces that the “functions” are not continuous
(since they are not defined) at those points. This is propagated as well into online
homework systems such as Webassign, etc.
Problem 6 Find the volume of the region bounded by the hyperbolic cylinders
xy = 1, xy = 9, xz = 4, xz = 36, yz = 25, yz = 49. Do only the part in the first
octant (x, y, z ≥ 0). [Hint: Make an appropriate change of variables and use the
change of variables theorem to compute it. ]
Setting up the problem and the change of variable theorem details are important
and then the integral computation (with the Jacobian)... I have directed students to
use change of variables to check whether they have learned this very process. Equal
points were allotted for substitution, for theorem conditions and for computation
of the integral/Jacobian.
Since each of the x, y, z combinations repeat twice in a circular fashion, the
substitutions seem to be imposing themselves: u = xy, v = xz, w = yz with bound-
aries for u|91 , v|36 49
4 , w|25 (*). Multiplying all the substitution equalities leads to
17
√
a helpful relationship: xyz = uvw (**). The map ψ(x, y, z) = (xy, xz, yz) =
(u(x, y, z), v(x, y, z), w(x, y, z)) transforms a “curvy, hyperbolic” x, y, z brick V into
a perpendicular u, v, w brick V0 [1, 9] × [4, 36] × [25, 49]. For ψ to be a good
substitution map, we need to check that it is one-to-one. That follows from
the fact that equalities x1 y1 = x2 y2 , x1 z1 = x2 z2 , y1 z1 = y2 z2 imply equalities
x1 = x2 , y1 = y2 , z1 = z2 . Thus existence of the inverse function ψ −1 = φ(u, v, w) =
= (x(u, v, w), y(u, v, w), z(u, v, w)) ispguaranteed.
p In p
fact, solving the substitution
equalities (*) gives φ(u, v, w) = ( uv/w, uw/v, vw/u), which is also one-
to-one.
√ The Jacobian is then computed to be ∂(u, v, w)/∂(x, y, z) = −2xyz =
−2 uvw ((**) was used). Both ψ, φ are C 1 maps, since all the partials exist and
they are continuous. Thus conditions for the change of variable formula are satisfied
and we find the volume as follows:
√
ZZZ ZZZ
dxdydz = | − 2 uvw|−1 dudvdw
V V0
9 36 49
1
Z Z Z
−1/2 −1/2
= u du v dv w−1/2 dw = 64.
2 1 4 25
There are a number of tripping points here, from routine to more subtle ones. Not
knowing/remembering how to compute the Jacobian (a fairly intuitive formula) was
a rare miss among the routine ones. The change of variables formula (a substitution
formula) is computed by
ZZZ
f (x, y, z)dxdydz =
V
ZZ
f (x(u, v, w), y(u, v, w), z(u, v, w))|Jac(φ)|dudvdw
V0
None of this is to be found in the student’s solution who had a lucky confounding
event with a formula for finding volume under a surface z (which would be a double
19
integral). Two more students start taking square roots of the given numbers, for
the temptation is too great not to do that when perfect squares of integers are
given.
It would help greatly teaching (and learning)the integration method of substi-
tution in several variables, if this method were thoroughly taught in the case of
one variable (usually in calculus I). The practice of applying the method of sub-
stitution in one variable almost exclusively omits checking that the substitutions
are of the right kind (namely that they are C 1 and one-to-one, etc.). Also, when
doing substitutions, students retain the old limits since they want to go back to the
old variable, once they eventually solve the indefinite integral via the substituting
variable. Thus an opportunity is missed to introduce (implicitly) the “Jacobian”
and all the other features that would become rather useful when the case of several
variables is taken up. I cannot overemphasize this point.
Problem 7. Find the volume of the smaller region bounded by the sphere
x + y 2 + z 2 = a2 (a is a given constant) and the plane z = b, where constant b is
2
eral substitution procedures and compute the Jacobians, but they certainly have
to know the spherical and cylindrical substitutions.
The next stage is to determine V0 , namely to determine integration limits for
the new variables. For this, well-drawn geometric diagrams are useful, and simple
appeal to the Pythagoras’
√ theorem will give
√ limits for cylindrical coordinates: 0 ≤
θ ≤ 2π, 0 ≤ r ≤ a2 − b2 and b ≤ z ≤ a2 − r 2 . For spherical coordinates, the
limits are as follows: 0 ≤ θ ≤ 2π, 0 ≤ φ ≤ arccos(b/a), and a ≤ ρ ≤ b/ cos φ.
Determining these limits was worth 1/3 of total points and so was the final part of
actually computing the integrals (and thus the volume): In cylindrical form
√ √
Z 2π Z a2 −b2 Z a2 −r2
V = dθ rdr dz =
0 0 b
√
Z a2 −b2 p
= 2π rdr( a2 − r 2 − b) =
0
√ √ !
Z a2 −b2 p Z a2 −b2
= 2π 2 2
r a − r dr − b rdr .
0 0
come partly from students confounding procedures they used to find areas under
curves (integrand is the function formula), or volumes under surfaces (integrand is
the surface formula with the double integral as a device). Thus these should be
contrasted with the volume formula when starting work with triple integrals.
With a number of errors along the way, students struggle through, coming up
with some answers that are more often than not polynomials in a, b (including nu-
merical constants). However, these polynomials range from being non-homogeneous,
21
One of the many uses of our analysis is in making lesson plans. Let us assume
that the lesson’s topic is continuity, one of the most important and most neglected
topics in calculus. We have feedback here attached to Problem 5 and we can
incorporate it into the lesson plan as follows:
Make sure that, by this time, students are well-familiar with the exact notion of
function, as an ordered triple – domain, codomain and the assignment rule(s). This
should be done at early pre-calculus, but in any case it is a must to review/teach at
the beginning of the course. Give students some “trick questions” such as ask them
to decide√which of the following are functions
√ and which are not: f : A −→ B,√a)
f (x) = x, A = R, B = R, b) f (x) = x, A = R+ , B = R, c) f (x) = − x,
22
√ √
A = R+ , B = R+ , d) f (x) = − x, A = R+ , B = R− , e) f (x) = x, A = R+ ,
B = R− . If complex numbers are part of the course, tailor this accordingly.
By this time enough exercises must have been done for students to feel comfort-
able with the notion of a limit and the routine of finding various limits, as well
as basic operations with limits. This should already be done in the first calculus
course. Test them on this with the following limits:
It is likely that the students will indiscriminately invoke “L’Hospital’s rule;” remind
them what conditions must be satisfied with this rule.
Now, use some limits, in higher dimensions, preferably those you will use to
check whether functions are continuous or not. First some that separate variables:
sin x cos y
lim ,
(x,y)→(0,π/2) xy
And some that are not obviously variable separating lim(x,y)→(a,b) [xy] (the inte-
ger part function). Discussion of cases when ab is and is not an integer may be
instructive.
Do also some piecewise defined function examples, for instance the example used
in Problem no. 5. Or the following one:
ex −1 sin y
x y , if x 6= 0 and y 6= 0,
sin y ,
if x = 0 and y 6= 0,
y
f (x, y) =
e −1
x
, if x 6= 0 and y = 0,
x
1, if x = 0 and y = 0.
Define continuity of function f at a point P ∈ A of the domain of the function f :
A −→ B if limx→P f (x) = f (P ). Emphasize that a question whether a “function”
is continuous at a point outside of its domain is meaningless, since there is no
function at such a point. Define what is meant to say that f is discontinuous at a
point. Define (global) continuity of f on all of its domain.
Now do first some simple examples of one variable, then pick up the aforemen-
tioned examples from limit exercises and add new ones. Make sure the examples
are graduated, from simple to more complicated.
If possible, visualize the functions and their limits, either by hand, or using
graphing software.
23
Conclusions
The problems with understanding functions exist from early mathematics educa-
tion and are dragged along throughout, including higher level or even graduate level
courses. I believe that the foundation for these problems begin before functions are
introduced, at the level of basic notions such as sets and ordered pairs (or triples,
etc). Graphing of functions is introduced simultaneously with the Cartesian plane,
without much ado about abstract formation of the Cartesian product of (two) sets.
Yet, it took many generations in human development before the idea of Cartesian
product matured. Not only should considerable time be devoted to “playing” with
coordinates, but a physical drawings6 of coordinates (and at least simple objects)
in two and three dimensional coordinate systems should be practiced thoroughly
until the students feel this as a “second nature.” Once this ground work is made
for the dimensions higher than one, one can introduce a name for a subset of a
Cartesian product A × B – call a subset ρ ⊆ A × B a relation from A to B, that can
be denoted by ρ : A −→ B.7 Some simple operations and drawings with relations
should be introduced as a practice.
Then, time comes for a definition of a function f : A −→ B as a relation with
properties that, for every a ∈ A there is only one b ∈ B with (a, b) ∈ f or b = f (a).
A should be called the domain of f and B the codomain and the range would
be introduced as the set of all b ∈ B such that (a, b) ∈ f , for all a ∈ A. An
important point should be made, namely that the function consists of these three
parts f : A −→ B and that each of them as important as any other in their own
right and that two functions f : A −→ B and g : C −→ D are equal if and only if
all three components are equal: A = C, B = D, f = g. A good number of exercises
that vary some or all parts of a function and discussion of resulting differences
should be made until students adopt this formalism to a high degree.
Then the stage is set for metaphors. A dynamic metaphor that I use is that of
a processor (a “grinding machine”) as sketched in the following diagram:
students are not intimidated by this abstraction level, which may run contrary to prejudices in
existence regarding mathematics education.
8 One can vary the description dependent on the fashionable language of the day and place –
iron bars that cannot be processed by the processor, or codomain that is a capillary
dish that cannot accommodate the processed dough... Varying all the parts of the
function would show a student that each part is in fact equally important as a
constituent of what is meant by a function. Composition may be explained in a
similar manner, for a packing machine g : C −→ D may be the follow up “factory”
that will work, only if the final product of f is in the domain of g; thus g may
be able to process sticks only of certain length range, not of other dimensions, yet
again underlining importance of domain and codomain (or range) – the diagram
has some sticks separated from the other in B, those in domain C of g and those
that are not in C.
There are other metaphors one can use, but a very important point to remem-
ber is that we cannot have correct metaphors if we do not give students correct
definitions of notions we teach them and insist that they adopt these definitions,
at least at the memorization level (first step in Bloom’s taxonomy). Otherwise we
would end up in finding metaphors, of metaphors, thus giving students an impres-
sion that the notions are negotiable moving targets. While mathematical formalism
is not equivalent to what mathematics is, it is nonetheless an essential constituent
of mathematics. It is most crucially visible in at least one practical development,
namely computerization of human knowledge and experience which could not take
place without strict mathematical formalism. Development of ideas is mostly a
meandering process, not infrequently through bouncing off of the extreme views or
practices. Thus an increased formalization of mathematics in the 20th century led
to sometimes indiscriminate transplantation of formalism into teaching practice,
which in turn led to overreaction in other direction, in denial of all the formalism
where the notions of function and its properties was degraded to pre-Eulerian levels
of understanding of what a function is. In a pivotal exposition of the pre-1930’s his-
tory of the development of function notion, Luzin (1935; see also Yushkevich (1966,
1977)) highlights subtleties of the function concept inclusive with the then develop-
ing views on classification of functions, intuitionistic developments, etc. Bourbaki
formalized the function concept in 1939 and then category theory brought about a
great generality and the accompanying simplification of the concept. Adoption of
these developments in school curriculum has been checkered at best and it would
be an interesting topic of research to look into the reasons why function concept is
not adopted and properly taught in schools (and colleges).
If this huge leap of students remembering correct definition of function gets
achieved, other constructs with functions would become much easier, one of them
being definition of continuity.
There are some drawbacks in analyzing or publishing an analysis of this kind.
For instance, it is a fact that analyzing tests in detail is very time consuming, and
even tedious. In addition, there are instances where this powerful educational tool
may be misused: Fearing negative feedback from students or administration alike,
some instructors would attempt to avoid potential trouble spots (as found through
analyses of this kind) by considerably diluting the level of content of the courses
25
Acknowledgment: The author would like to thank Professor Annie Selden for
her useful feedback on the content and form of the paper, and Professor Ivko Dim-
itrić for their stylistic comments, as well as anonymous referees for their comments
that led to a better presentation of the content of this paper.
References
Anderson, L.W. & Krathwohl, D.R. & Airasian, P.W. & Cruikshank, K.A. &
Mayer, R.E. & Pintrich, P.R. & Raths, J. & Wittrock, M.C. (Eds.) (2001): A
taxonomy for learning, teaching, and assessing. A revision of Bloom’s taxonomy of
educational objectives; Addison Wesley Longman, Inc.
Bloom, B.S. (Ed.) (1956). Taxonomy of Educational Objectives: The Classifi-
cation of Educational Goals, pp.201–207; Susan Fauer Company, Inc.
Breindenbach, Daniel & Dubinsky, Ed & Hawks, Julie & Nichols, Devilyna
(1992): Development of the process conception of function. Educational studies
in mathematics 23, 247–285.
Dimitric, R.M. (2001): Using less calcus in teaching calculus. A historical ap-
proach. Mathematics Magazine 74, No.3 (June), 201-211.
Dimitric, R.M. (2003): Components of Successful Education, Tenth Interna-
tional Congress of Mathematics Education (ICME10), TSG 27, 2004; The Teaching
of Mathematics, 6(2003), No.2, 69–80
Dimitric, R.M. (2004): Is 1/x continuous at 0?, arXiv: 1210.2939v1;mathGM]
Preprint.
Dimitric, R.M. (2009): Student placement in calculus courses. A case study.
Eleventh International Congress of Mathematics Education (ICME11), TSG , 2009;
The Teaching of Mathematics, 12(2009), No. 2, 83-102.
Dimitric, R.M. (2012a): How to make a course specific diagnostic test in the
context of Bloom’s taxonomy, In preparation.
Dimitric, R.M. (2012b): On complexity distance and its applications in educa-
tion, In preparation.
Dubinski, E. & Schoenfeld, A.H. & Kaput, J.J. (Eds) (1994): Research in Col-
legiate Mathematic Education, I. Providence, RI: American Mathematical Society.
Dubinski, E. & Harel, G. (1992): The nature of the process conception of func-
tion. In G. Harel & E. Dubinsky (Eds.). The concept of function: Aspects of
epistemology and pedagogy. MAA Notes 25, 85–106.
Horscovics, N. (1982): Problems related to the understanding of functions, in
Proceedings of the Workshop on Functions, Organized by the Foundation for Cur-
riculum Development, Enschede.
27
Leinhardt, G. & Zaslavsky, O. & Stein, M. (1990): Functions, graphs and graph-
ing: Tasks, learning and teaching, Review of Educational Research 60(1), 1-64.
Luzin, N.N. (1935): Function, The Great Soviet Encyclopaedia 1959, 314-334
Translated by Abe Shenitzer: Function: Part I Amer. Math Monthly 105, No. 1
(Jan. 1998), 59–67 and No. 3 (Mar. 1998), 263–270).
Oehrtman, Michael & Carlson, Marylin & Thomson, Patrick W. (2008): Foun-
dational reasoning abilities that promote coherence in students’ function under-
standing, In: Making the Connection: Research and Teaching in Undergraduate
Mathematics, SIGMA and RUME, MAA Notes, Vol. 75, 27–41.
Piaget, J. & Grize, J.-B. & Szeminska, A. & Bang, V. (1977): Epistemology and
Psychology of Functions (J. castellanos and V. Anderson, trans.), Reidel, Dordrecht.
Rogawski, J. (2008): Calculus. Early Transcendentals, 3rd printing. New York:
W.H. Freeman and Company.
Selden, J. & Mason, A. & Selden, A. (1989): Can average calculus students solve
nonroutine problems? Journal of Mathematical Behavior 8, 45–50.
Selden, J. & Selden, A. & Mason, A. (1994): Even good calculus students
can’t solve nonroutine problems. IN: Research issues in undergraduate mathemat-
ics learning (Vol. 33, pp.19-26). Washington, D.C.: Mathematical Association of
America.
Selden, Annie & Selden, John & Hauk, Shandy & Mason, Alice(2010): Why
Can’t Calculus Students Access their Knowledge to Solve Nonroutine Problems?
CBMS Issues in Mathematics Education, 8, 128–153.
Stewart, J. (2012): Single Variable Calculus. Early Transcendentals. 7th ed.
Australia: Brooks/Cole, Cengage Learning.
Sfard, A. (1987): Two conceptions of mathematical notions: Operational and
structural, in J.C. Bergeron, N. Herscovics and C. Kieran (eds.), Proceedings of the
Eleventh International Conference for the Psychology of Mathematics Education,
III, Université, Montreal, pp 162-169.
Tall, D. & Vinner, S. (1981): Concept images and concept definitions in math-
ematics with particular reference to limits and continuity, Educational Studies in
Mathematics 12, 151–169.
Thomson, Patrick W. (1994): Students, functions, and the undergradute cur-
riculum. In Dubinsky et al (Issues in Mathematics Education Vol 4, pp 21-24).
Vinner, S. (1989): Images and definition for the concept of function, Journal for
Research in Mathematics Education 20, 356–366.
Vinner, S. (1992): The function concept as a prototype for problems in mathe-
matics learning, in Dubinsky & Harel (1992), 195–214.
Vinner, S. & Dreyfus, T. (1989): Images and definitions for the concept of
function, Journal of Research in Mathematics Education 20. 356–366.
Yushkevich, A.P. (1966): The concept of function up to the middle of the nine-
teenth century, Istor.-Mat. Issled. , 17, 123–150.
Yushkevich, A.P. (1977): The concept of function up to the middle of the nine-
teenth century, Arch. History of Exact Sci. , 16 37–85.
28
What was the last math course you took, when and where?
Career goal (what is the reason you are taking this course):
Please do the following problems in one hour, without using a calculator. These
problems are a quick sample of calculus and algebra you will need for Calculus III.
I want to see how fresh you are with the material and advise you accordingly.
x + y = 1, x3 + y 3 = 2.
sin 3x
lim =
x→0 5x
11) Write the equation of the tangent line to the graph of the function
x
f (x) = ,
1 + x2
2
17) If f (x) = ae2x , what is f (2a) ? (a is a given constant).
18) Calculate the integral
1
x
Z
dx =
0 1 + x2
R∞ 2
19) Calculate 1
xe−x dx =
20) Calculate
Z π/4
tan x dx =
0